A Lean Approach for Evolving Heterogeneous Wireless Sensor Networks An Assisted Living Case Study Thomas Patzke [email protected] Software.

A Lean Approach for Evolving

Heterogeneous Wireless Sensor Networks

An Assisted Living Case Study

Thomas [email protected] Product Line Department

Lórant Vajda, Attila Török{vajda,torok}@ikti.huInstitute for Applied Telecommunication Technologies

RCEAS 2007Budapest, November 23

© Fraunhofer 2007 RCEAS 2007Budapest,

A Lean Approach for Evolving Heterogeneous Wireless

Sensor Networks

November 2007

Overview

Context & Problems

Consequences

Solution Ideas

Product Lines

Case Study

Lessons Learnt

© Fraunhofer 2007 RCEAS 2007Budapest,

A Lean Approach for Evolving Heterogeneous Wireless

Sensor Networks

November 2007

Context & Problems

Bilateral German-Hungarian Collaboration Project on Ambient Intelligence Systems (2005-08)

Subproject: Ambient Assisted Living

Application problem: Supporting elderly people with AmI technology to live longer in their own home

Engineering problems:

• building and

• evolving („maintaining“)

• high-quality AmI systems

• in a cost-effective way

© Fraunhofer 2007 RCEAS 2007Budapest,

A Lean Approach for Evolving Heterogeneous Wireless

Sensor Networks

November 2007

Consequences

Evolution aspects:

• Space- heterogeneous systems

• Time- system extension or contraction

Quality & cost aspects:

• At runtime- Functionality, efficiency (little resources!), safety

• During construction- Minimal construction & evolution effort

- Rapid response to changes

© Fraunhofer 2007 RCEAS 2007Budapest,

A Lean Approach for Evolving Heterogeneous Wireless

Sensor Networks

November 2007

Solution Ideas

Reuse: WSN systems as a product line (HW & SW)

• benefit from their similarities

• across the entire system engineering life-cycle

- requirements, analysis, design, implementation

Simplicity

• by feature prioritization

• by removing arbitrary complexities

• goal: simple-enough systems

Vision: Self-generating autonomous systems

© Fraunhofer 2007 RCEAS 2007Budapest,

A Lean Approach for Evolving Heterogeneous Wireless

Sensor Networks

November 2007

Large-Scale-Reuse Approach: Product Lines

A (software) PL is a set of (SW) products that are developed and evolved together

Product

1Product

2Product

3

Single systems

1

2

3

Product line

Important PL concepts• Commonality - what PL members have in common• Variability - where PL members differ• Variation points - where the variation occurs• Defaults - what most (but not all) PL members share

© Fraunhofer 2007 RCEAS 2007Budapest,

A Lean Approach for Evolving Heterogeneous Wireless

Sensor Networks

November 2007

WSNs as a Product Line – The Feature Model

WSN

Sensor Nodes

Send Receive

Legend:

Mandatory feature

Optional feature

Broadcast

Unicast

Sense Actuate

Adapter Nodes

Sink Nodes

© Fraunhofer 2007 RCEAS 2007Budapest,

A Lean Approach for Evolving Heterogeneous Wireless

Sensor Networks

November 2007

Case Study – Heterogeneous Technologies

MicaZ and Particle Computer

Domain analysis results:

• similar kinds of sensors (acceleration, light, temperature, sound)

• similar transmission modality (send & receive, wirelessly)

• but: different communication modes (pull vs. push)

• different node programming language (nesC vs. C)

• same sink programming language (Java)

© Fraunhofer 2007 RCEAS 2007Budapest,

A Lean Approach for Evolving Heterogeneous Wireless

Sensor Networks

November 2007

Case Study – Integration

Experiment: Combine the „technology clusters“ in all different send/receive combinations, and in gathering redundant data

Possibilities:

• Direct hardware connection

- combinatory explosion of connections

• Common adapter/gateway

- localizes exchange

- chosen because of simplest integration

Integration of Push (Event) & Pull (Poll) Approach

© Fraunhofer 2007 RCEAS 2007Budapest,

A Lean Approach for Evolving Heterogeneous Wireless

Sensor Networks

November 2007

Lessons Learnt

Even using the same programming language (Java) does not warrant seamless integration (version differences!)

Text (as opposed to binary data) is a simple, valuable format for distributing aggregated sensor data

• facilitates development and evolution- avoids data size & endianness problems

- is human-readable

• promotes common metadata formats (CSV,…)

- easily reusable across technology clusters

It is not wise to focus on efficiency first

© Fraunhofer 2007 RCEAS 2007Budapest,

A Lean Approach for Evolving Heterogeneous Wireless

Sensor Networks

November 2007

Further Information

BelAmI project

• www.belami-project.org

Product Lines

• www.iese.fhg.de/fhg/iese/research/development/pla

• www.softwareproductlines.com

Institute for Applied Telecommunication Technologies

• www.ikti.hu